结合自适应暗通道先验和图像融合策略的单幅图像除雾方法

为解决暗通道先验统计学模型在一些情况下存在“光晕效应”、颜色偏暗和在雾浓度高区域处理效果不佳等问题,针对暗通道先验方法进行改进,并结合图像融合策略来增强可视化区域的视觉效果.利用像素块加权插值法来计算每个像素点的暗通道值,进而消除软抠图或导向滤波方法所带来的光晕效应;利用高斯模型对待恢复图像的暗通道像素值进行模拟,从而自适应地恢复天空和其他明亮区域;通过图像融合策略增强高浓度区域的图像信息.实验结果表明,与其他几种经典算法相比,改进方案不仅能够显著提高有雾图像的可见度,而且具有更好的鲁棒性.

Single image haze removal using adaptive dark channel prior and image fusion strategy

To resolve the problems of over-saturation, artefacts and dark-look for Dark Channel Prior, this study proposes a method of single image haze removal using adaptive dark channel and image fusion. The Weighted Normalization Interpolation Method is used to compute the dark channel of pixel. The dark channel pixels of the image going to be recovered are modeled as a Gaussian one that a more natural recovered image of the sky and other bright regions can be obtained adaptively. Finally, a post fusion method is devised to increase the image information at dense haze region. Experimental results demonstrate that the proposed method not only significantly improves the visibility of the hazy image than the well-known state-of-the-art approaches, but also has a better robustness.